Using Deep Neural Networks To Classify Coral Reef And Algae

School Name

South Carolina Governor's School for Science & Mathematics

Grade Level

12th Grade

Presentation Topic

Computer Science

Presentation Type

Mentored

Written Paper Award

1st Place

Abstract

The human brain is easily the most complex structure in the known universe. Its processing power completely dwarfs the capabilities of the modern computer and most likely will for a long time. Even with all that processing power, however, computers are still far more accurate than humans. For this reason, the study of Machine Learning has risen dramatically. Machine Learning is the study of creating advance networks of mathematical computations that allow computers to “learn” how to classify and identify different problems, such as classifying dogs vs cats with high levels of accuracy. This lab is using the studies from Machine Learning to create a Convolutional Neural Network to process images of coral reefs that are feed to it through an aquatic robot out in the field. The computer can, with minimal effort, identify the different types of coral reefs with an accuracy of around 40%. The difficulty of this experiment is in the fine tuning of the Neural Network to allow an accuracy of around 95% and higher. Over the summer, we spent our time creating such a network that could classify coral reef. This model, if it were to be completed, would have been able to be used in tandem with the aquatic robots so that they could use the coral reef as landmarks to create virtual maps of underwater areas. They would have been able to use the virtual maps to plot out paths for navigation autonomously rather than human plotting paths for them.

Location

Founders Hall 140 A

Start Date

3-30-2019 11:30 AM

Presentation Format

Oral Only

Group Project

No

COinS
 
Mar 30th, 11:30 AM

Using Deep Neural Networks To Classify Coral Reef And Algae

Founders Hall 140 A

The human brain is easily the most complex structure in the known universe. Its processing power completely dwarfs the capabilities of the modern computer and most likely will for a long time. Even with all that processing power, however, computers are still far more accurate than humans. For this reason, the study of Machine Learning has risen dramatically. Machine Learning is the study of creating advance networks of mathematical computations that allow computers to “learn” how to classify and identify different problems, such as classifying dogs vs cats with high levels of accuracy. This lab is using the studies from Machine Learning to create a Convolutional Neural Network to process images of coral reefs that are feed to it through an aquatic robot out in the field. The computer can, with minimal effort, identify the different types of coral reefs with an accuracy of around 40%. The difficulty of this experiment is in the fine tuning of the Neural Network to allow an accuracy of around 95% and higher. Over the summer, we spent our time creating such a network that could classify coral reef. This model, if it were to be completed, would have been able to be used in tandem with the aquatic robots so that they could use the coral reef as landmarks to create virtual maps of underwater areas. They would have been able to use the virtual maps to plot out paths for navigation autonomously rather than human plotting paths for them.